Proc. Nati. Acad. Sci. USA Vol. 75, No. 10, pp 4848-4852, October 1978 Bfiochemistry
Glucose-induced conformational change in yeast hexokinase (protein crystallography/induced fit/interdomain protein flexibility/hydrophobic effect)
WILLIAM S. BENNETT, JR., AND THOMAS A. STEITZ Department of Molecular Biophysics and Biochemistry, Yale University, New Haven, Connecticut 06520
Communicated by Julian M. Sturtevant, July 31, 1978
ABSTRACT The A isozyme of yeast hexokinase (ATP:Dhexose 6-phosphotransferase, EC 2.7.1.1) crystallized as a complex with glucose has a conformation that is dramatically different from the conformation of the B isozyme crystallized in the absence of glucose. Comparison of the high-resolution structures shows that one lobe of the molecule is rotated by 120 relative to the other lobe, resulting in movements of as much as 8 A in the polypeptide backbone and closing the cleft between the lobes into which glucose is bound. The conformational change is produced by the binding of glucose (R. C. McDonald, T. A. Steitz, and D. M. Engelman, unpublished data) and is essential for catalysis [Anderson, C. M., Stenkamp, R. E., McDonald,RIC.& Steitz, T. A. (1978) J. Mol. Biol. 123, 207-2191Jand thus provides an example of induced fit. The surface area of the hexokcinase Aeglucose complex exposed to solvent is smaller than that of native hexokinase B. By using the change in exposed surface area to estimate the hydrophobic contribution to the free energy changes upon glucose binding, welfind that the hydrophobic effect alone favors the active conformation of hexokinase in the presence and absence of sugar. The observed stability of the inactive conformation of the enzyme in the absence of substrates may result from a deficiency of complementary interactions within the cavity that forms when the two lobes close together. The ability of hexokinase (ATP:D-hexose 6-phosphotransferase, EC 2.7.1.1) to discriminate against water as a substrate led Koshland (1) to suggest that this enzyme utilizes an "induced fit" mechanism of enzyme specificity. He proposed that sugar substrates can induce a protein conformational change that is essential for catalysis, whereas water cannot provide the specific interactions with the protein that are required to induce the change. Considerable experimental support exists for induced fit in yeast hexokinase. Kaji and Colowick (2) detected a small ATPase activity in the absence of sugar and showed that the Km for ATP in this reaction was 50 times higher than in the hexokinase reaction. DelaFuente et al. (3) showed that the ATPase activity could be stimulated by hexose analogs lacking the reactive 6-hydroxymethyl group, suggesting that the unreactive pyranose ring of these analogs could produce the required conformational change. More direct evidence that sugars actually induce some conformational change in the enzyme was provided by earlier crystallographic studies. High concentrations of glucose shatter all crystal forms of the B isozyme (W. F. Anderson and T. A. Steitz, unpublished observations; ref. 4) and, at lower concentrations, produce small structural changes throughout one lobe of the enzyme (5, 6). Other kinetic and spectroscopic evidence consistent with a glucose-induced conformational change has also been reported (7-9). We present here the structure of a complex between yeast hexokinase A and glucose (HKA-G) at 3.5-A resolution. Comparison of this structure with the 2.1-A resolution structure of the native hexokinase B monomer (BIII; refs. 10 and 11) reveals
a substantial difference in structure that has been identified with the glucose-induced conformational change shown to be essential for catalysis (12). Using these two structures, we have examined the roles of hydrophobic forces and structural complementarity in this ligand-induced conformational change.
The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U. S. C. §1734 solely to indicate this fact.
Abbreviations: E and E', inactive and active enzyme; G, glucose; HKA-G, hexokinase A complexed with glucose; BIII, crystal form BIll of hexokinase B; rms, root mean square; OTG, o-toluoyl-2-glucos-
EXPERIMENTAL Crystallization of HKA-G Complex. Crystals of yeast HKA-G complex were grown at room temperature by a method similar to that of Womack et al. (13): protein (5 to 10 mg/ml) was dialyzed against 50-55% saturated ammonium sulfate/0.1 M potassium phosphate, pH 6.6/0.01 M glucose. Glucose is strictly required for crystal growth and preservation. Crystals are in space group P212121 (a = 145.0 A, b = 79.0 A, c = 62.0 A) with one monomer per asymmetric unit. All hexokinase A used in these studies was the generous gift of F. C. Womack and S. P. Colowick of Vanderbilt University. Structure Determination. The HKA-G structure was independently solved at 4.5-A resolution by the multiple isomorphous replacement method using two heavy atom derivatives, platinum(II) chloride and potassium osmiate. The mean figure of merit was 0.68. Data measurement and processing procedures were similar to those described by Anderson et al. (10). Because this electron density map supported our earlier conclusion (14) that the tertiary structures of the hexokinase isozymes were very similar except for a change in the relative orientation of two domains, we used the refined BIII model (11) to extend the phasing of HKA.G data to 3.5 A resolution. The BIII model was positioned in the HKA-G unit cell (14),.and the orientation of the small lobe was adjusted by hand to fit the HKA-G electron density. The orientation of each lobe was then refined independently to optimize its fit to the HKA-G electron density by the real-space procedure of Fletterick and Wyckoff (15). The resulting HKA-G model was subjected to several cycles of difference Fourier refinement alternated with idealization of the model, using procedures similar to those of Anderson et al. (11). Because the amino acid sequence of the A isozyme is not known, the sequence of the B isozyme obtained from the x-ray structure (11) was used. This sequence was not changed during the HKA-G refinement. There are several indications that the HKA-G structure is essentially correct. Refinement of the HKA-G model was terminated when further progress would have required altering the x-ray-determined (11) amino acid sequence. The final (FQ - Fc) difference electron density map (in which F. and Fc are the observed and calculated structure factors) showed fewer features than the initial difference maps and the conventional crystallographic R factor for the final model was 0.26 at 3.5-A resolution. The root mean square (rms) deviation of bond
amine.
4848
Biochemistry:
Bennett and Steitz
Proc. Natl. Acad. Sci. USA 75 (1978)
4849
FIG. 1. Stereo drawing of the C, backbone of the HKA-G and Bill models after the large lobes were superimposed as described in the legend to Fig. 2. HKA.G is represented with solid bonds between adjacent Ca, atoms and the small lobe of BIII with dashed bonds. The small lobe as drawn contains 37% of the atoms of the molecule. Because the rest of the BIll structure is similar to that of HKA-G (Fig. 2), it is omitted for clarity. The atomic coordinates of glucose (11) and ATP (6) from binding experiments on the BIII crystal form are represented by open circles. The long line piercing the left end of the molecule indicates the position of the screw axis relating the two small lobes.
lengths from ideality was 0.1 A. Heavy atom or substrate difference electron density maps calculated using phases derived from the model improved markedly during the refinement. Also, (FO - F) difference maps using phases and structure factors calculated from a model in which a portion of the refined protein was omitted clearly showed the density of the deleted region.* We have estimated the rms coordinate error to be 0.45 A by the method of Luzzati (16). Although this model may be inaccurate in some places, especially where side chains differ between isozymes, we have considerable confidence in the placement of the polypeptide backbone. Details of the structure determination will be described elsewhere. * Solvent Accessibility Calculations. The areas of the HKA-G and BIII structures accessible to solvent were calculated by the procedure of Lee and Richards (17) by using computer programs written by T. J. Richmond of Yale University. To compare the magnitude of the hydrophobic effect resulting from changes in exposed surface area with the corresponding equilibria of hexokinase in solution, hydrophobic free energy changes for the glucose-induced conformational change were estimated from changes in accessible surface area by using the empirical relationship developed by Chothia (18). Because a hydrogen-bonded donor-acceptor pair behaves essentially as a nonpolar group in protein packing (19) and most of the hydroxyl groups of a glucose bound to hexokinase are extensively hydrogen bonded (12), we have used the same relationship between the accessible surface area of glucose and hydrophobic free energy in these estimates. The hydrogen bonds made by the bound glucose are assumed in our calculations to be equivalent to those made with water and to provide no net binding energy. RESULTS The Conformational Change. Comparison of the refined HKA G and BIII models confirms that their tertiary structures are nearly identical except for a large change in the relative orientations of two lobes. To demonstrate this change, the C, atoms of each of the lobes of the two models were superimposed by using a least-squares procedure, treating the lobes as rigid bodies. The large lobes (residues 2-58 and 187-458) of the HKA-G and BIII models superimpose with an rms residual of 1.0 A between C0 atoms. The small lobes (residues 59-186) superimpose with an rms residual of 1.6 A. Although there are small systematic deviations from the rigid-body fit near crystal packing contacts and in some regions of the small lobe that are brought into contact with glucose,* the low residuals of the *
W. S. Bennett and T. A. Steitz, unpublished.
superposition show that each lobe behaves predominantly as a rigid body in the conformational change between the native and glucose-complex structures. Fig. 1 shows the positions of the small lobe in the two structures after superposition of the large lobes, and Fig. 2 shows the distance between equivalent C0 atoms with the structures aligned as in Fig. 1. Parts of the polypeptide backbone of the small lobe move as much as 8 A towards the large lobe in the glucose complex, closing the cleft in which the glucose molecule is bound and bringing atoms of the small lobe into contact with both the large lobe and the substrates. The conformational change may also be described by the rigid-body transformation needed to superimpose the small lobes once the large lobes are aligned as in Fig. 1. The resulting transform, expressed as a single screw operation by the method of Cox (21), corresponds to a rotation of 120 around and a translation of 0.9 A along the screw axis shown in the figure. Surface Area Changes. Table 1 presents the results of comparing accessible surface areas of the hexokinase species represented schematically in Fig. 3 and compares the equilibrium constants calculated from this hydrophobic effect alone
7-
6
-
a5C
13
2~~~~~~~~~~~~2 0
1 00
200
300
400
500
Residue
FIG. 2. Magnitude of the conformational change. The distance between equivalent C,, atoms is plotted after superposition of the large lobes of the HKA-G and Bill models by a rigid-body transformation giving the best least-squares fit of selected C,, atoms. Short lengths of polypeptide whose conformations are different in the two large lobes were systematically omitted from the least-squares procedure by criteria similar to those of Rossmann and Argos (20). These omissions amount to 24% of the C
Glucose-induced conformational change in yeast hexokinase (protein crystallography/induced fit/interdomain protein flexibility/hydrophobic effect)
WILLIAM S. BENNETT, JR., AND THOMAS A. STEITZ Department of Molecular Biophysics and Biochemistry, Yale University, New Haven, Connecticut 06520
Communicated by Julian M. Sturtevant, July 31, 1978
ABSTRACT The A isozyme of yeast hexokinase (ATP:Dhexose 6-phosphotransferase, EC 2.7.1.1) crystallized as a complex with glucose has a conformation that is dramatically different from the conformation of the B isozyme crystallized in the absence of glucose. Comparison of the high-resolution structures shows that one lobe of the molecule is rotated by 120 relative to the other lobe, resulting in movements of as much as 8 A in the polypeptide backbone and closing the cleft between the lobes into which glucose is bound. The conformational change is produced by the binding of glucose (R. C. McDonald, T. A. Steitz, and D. M. Engelman, unpublished data) and is essential for catalysis [Anderson, C. M., Stenkamp, R. E., McDonald,RIC.& Steitz, T. A. (1978) J. Mol. Biol. 123, 207-2191Jand thus provides an example of induced fit. The surface area of the hexokcinase Aeglucose complex exposed to solvent is smaller than that of native hexokinase B. By using the change in exposed surface area to estimate the hydrophobic contribution to the free energy changes upon glucose binding, welfind that the hydrophobic effect alone favors the active conformation of hexokinase in the presence and absence of sugar. The observed stability of the inactive conformation of the enzyme in the absence of substrates may result from a deficiency of complementary interactions within the cavity that forms when the two lobes close together. The ability of hexokinase (ATP:D-hexose 6-phosphotransferase, EC 2.7.1.1) to discriminate against water as a substrate led Koshland (1) to suggest that this enzyme utilizes an "induced fit" mechanism of enzyme specificity. He proposed that sugar substrates can induce a protein conformational change that is essential for catalysis, whereas water cannot provide the specific interactions with the protein that are required to induce the change. Considerable experimental support exists for induced fit in yeast hexokinase. Kaji and Colowick (2) detected a small ATPase activity in the absence of sugar and showed that the Km for ATP in this reaction was 50 times higher than in the hexokinase reaction. DelaFuente et al. (3) showed that the ATPase activity could be stimulated by hexose analogs lacking the reactive 6-hydroxymethyl group, suggesting that the unreactive pyranose ring of these analogs could produce the required conformational change. More direct evidence that sugars actually induce some conformational change in the enzyme was provided by earlier crystallographic studies. High concentrations of glucose shatter all crystal forms of the B isozyme (W. F. Anderson and T. A. Steitz, unpublished observations; ref. 4) and, at lower concentrations, produce small structural changes throughout one lobe of the enzyme (5, 6). Other kinetic and spectroscopic evidence consistent with a glucose-induced conformational change has also been reported (7-9). We present here the structure of a complex between yeast hexokinase A and glucose (HKA-G) at 3.5-A resolution. Comparison of this structure with the 2.1-A resolution structure of the native hexokinase B monomer (BIII; refs. 10 and 11) reveals
a substantial difference in structure that has been identified with the glucose-induced conformational change shown to be essential for catalysis (12). Using these two structures, we have examined the roles of hydrophobic forces and structural complementarity in this ligand-induced conformational change.
The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U. S. C. §1734 solely to indicate this fact.
Abbreviations: E and E', inactive and active enzyme; G, glucose; HKA-G, hexokinase A complexed with glucose; BIII, crystal form BIll of hexokinase B; rms, root mean square; OTG, o-toluoyl-2-glucos-
EXPERIMENTAL Crystallization of HKA-G Complex. Crystals of yeast HKA-G complex were grown at room temperature by a method similar to that of Womack et al. (13): protein (5 to 10 mg/ml) was dialyzed against 50-55% saturated ammonium sulfate/0.1 M potassium phosphate, pH 6.6/0.01 M glucose. Glucose is strictly required for crystal growth and preservation. Crystals are in space group P212121 (a = 145.0 A, b = 79.0 A, c = 62.0 A) with one monomer per asymmetric unit. All hexokinase A used in these studies was the generous gift of F. C. Womack and S. P. Colowick of Vanderbilt University. Structure Determination. The HKA-G structure was independently solved at 4.5-A resolution by the multiple isomorphous replacement method using two heavy atom derivatives, platinum(II) chloride and potassium osmiate. The mean figure of merit was 0.68. Data measurement and processing procedures were similar to those described by Anderson et al. (10). Because this electron density map supported our earlier conclusion (14) that the tertiary structures of the hexokinase isozymes were very similar except for a change in the relative orientation of two domains, we used the refined BIII model (11) to extend the phasing of HKA.G data to 3.5 A resolution. The BIII model was positioned in the HKA-G unit cell (14),.and the orientation of the small lobe was adjusted by hand to fit the HKA-G electron density. The orientation of each lobe was then refined independently to optimize its fit to the HKA-G electron density by the real-space procedure of Fletterick and Wyckoff (15). The resulting HKA-G model was subjected to several cycles of difference Fourier refinement alternated with idealization of the model, using procedures similar to those of Anderson et al. (11). Because the amino acid sequence of the A isozyme is not known, the sequence of the B isozyme obtained from the x-ray structure (11) was used. This sequence was not changed during the HKA-G refinement. There are several indications that the HKA-G structure is essentially correct. Refinement of the HKA-G model was terminated when further progress would have required altering the x-ray-determined (11) amino acid sequence. The final (FQ - Fc) difference electron density map (in which F. and Fc are the observed and calculated structure factors) showed fewer features than the initial difference maps and the conventional crystallographic R factor for the final model was 0.26 at 3.5-A resolution. The root mean square (rms) deviation of bond
amine.
4848
Biochemistry:
Bennett and Steitz
Proc. Natl. Acad. Sci. USA 75 (1978)
4849
FIG. 1. Stereo drawing of the C, backbone of the HKA-G and Bill models after the large lobes were superimposed as described in the legend to Fig. 2. HKA.G is represented with solid bonds between adjacent Ca, atoms and the small lobe of BIII with dashed bonds. The small lobe as drawn contains 37% of the atoms of the molecule. Because the rest of the BIll structure is similar to that of HKA-G (Fig. 2), it is omitted for clarity. The atomic coordinates of glucose (11) and ATP (6) from binding experiments on the BIII crystal form are represented by open circles. The long line piercing the left end of the molecule indicates the position of the screw axis relating the two small lobes.
lengths from ideality was 0.1 A. Heavy atom or substrate difference electron density maps calculated using phases derived from the model improved markedly during the refinement. Also, (FO - F) difference maps using phases and structure factors calculated from a model in which a portion of the refined protein was omitted clearly showed the density of the deleted region.* We have estimated the rms coordinate error to be 0.45 A by the method of Luzzati (16). Although this model may be inaccurate in some places, especially where side chains differ between isozymes, we have considerable confidence in the placement of the polypeptide backbone. Details of the structure determination will be described elsewhere. * Solvent Accessibility Calculations. The areas of the HKA-G and BIII structures accessible to solvent were calculated by the procedure of Lee and Richards (17) by using computer programs written by T. J. Richmond of Yale University. To compare the magnitude of the hydrophobic effect resulting from changes in exposed surface area with the corresponding equilibria of hexokinase in solution, hydrophobic free energy changes for the glucose-induced conformational change were estimated from changes in accessible surface area by using the empirical relationship developed by Chothia (18). Because a hydrogen-bonded donor-acceptor pair behaves essentially as a nonpolar group in protein packing (19) and most of the hydroxyl groups of a glucose bound to hexokinase are extensively hydrogen bonded (12), we have used the same relationship between the accessible surface area of glucose and hydrophobic free energy in these estimates. The hydrogen bonds made by the bound glucose are assumed in our calculations to be equivalent to those made with water and to provide no net binding energy. RESULTS The Conformational Change. Comparison of the refined HKA G and BIII models confirms that their tertiary structures are nearly identical except for a large change in the relative orientations of two lobes. To demonstrate this change, the C, atoms of each of the lobes of the two models were superimposed by using a least-squares procedure, treating the lobes as rigid bodies. The large lobes (residues 2-58 and 187-458) of the HKA-G and BIII models superimpose with an rms residual of 1.0 A between C0 atoms. The small lobes (residues 59-186) superimpose with an rms residual of 1.6 A. Although there are small systematic deviations from the rigid-body fit near crystal packing contacts and in some regions of the small lobe that are brought into contact with glucose,* the low residuals of the *
W. S. Bennett and T. A. Steitz, unpublished.
superposition show that each lobe behaves predominantly as a rigid body in the conformational change between the native and glucose-complex structures. Fig. 1 shows the positions of the small lobe in the two structures after superposition of the large lobes, and Fig. 2 shows the distance between equivalent C0 atoms with the structures aligned as in Fig. 1. Parts of the polypeptide backbone of the small lobe move as much as 8 A towards the large lobe in the glucose complex, closing the cleft in which the glucose molecule is bound and bringing atoms of the small lobe into contact with both the large lobe and the substrates. The conformational change may also be described by the rigid-body transformation needed to superimpose the small lobes once the large lobes are aligned as in Fig. 1. The resulting transform, expressed as a single screw operation by the method of Cox (21), corresponds to a rotation of 120 around and a translation of 0.9 A along the screw axis shown in the figure. Surface Area Changes. Table 1 presents the results of comparing accessible surface areas of the hexokinase species represented schematically in Fig. 3 and compares the equilibrium constants calculated from this hydrophobic effect alone
7-
6
-
a5C
13
2~~~~~~~~~~~~2 0
1 00
200
300
400
500
Residue
FIG. 2. Magnitude of the conformational change. The distance between equivalent C,, atoms is plotted after superposition of the large lobes of the HKA-G and Bill models by a rigid-body transformation giving the best least-squares fit of selected C,, atoms. Short lengths of polypeptide whose conformations are different in the two large lobes were systematically omitted from the least-squares procedure by criteria similar to those of Rossmann and Argos (20). These omissions amount to 24% of the C
